fix: reject ties without draw probability; never report NaN as converged

A tie with `p_draw == 0.0` produced NaN posteriors in release builds and
`converge()` reported `converged: true`, because every comparison against
NaN is false and `tuple_gt` therefore read NaN as "below epsilon".

Two independent defects, fixed together:

- Ingestion now rejects tied outcomes when the draw probability is zero,
  promoting the existing `debug_assert!` in `Game::ranked_with_arena` to a
  real `InferenceError::TieWithoutDrawProbability`. Validation sits in
  `add_events_with_prior`, the chokepoint every route reaches — including
  `record_draw`, which bypasses `Outcome` entirely.
- `converge()` treats a non-finite step as failure and returns
  `InferenceError::NonFiniteResult` rather than claiming convergence.

Also in this change:

- `History::converge()` on an empty history returned a `usize` underflow
  panic from `0..len()-1`; it now short-circuits to a zero-iteration report.
- `Outcome::scores_with_sigma` no longer panics on a non-positive sigma;
  the value is validated at ingestion so callers get an error instead.
- `InferenceError` gains `WrongOutcomeKind`, replacing the misuse of
  `MismatchedShape` for variant mismatches (which rendered as the nonsense
  "expected length 0, got 0"), and is now `#[non_exhaustive]`.

Note `Outcome::winner(w, n)` for n >= 3 ties every loser, so those events
now require a positive `p_draw`. They previously returned NaN.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DnsaJg74eNSva3PJjK2eej
This commit is contained in:
2026-08-04 21:38:29 +02:00
co-authored by Claude Opus 5
parent 2b5d3b1687
commit f4e2922d59
6 changed files with 371 additions and 19 deletions
+41
View File
@@ -1,6 +1,7 @@
use std::fmt;
#[derive(Debug, Clone, PartialEq)]
#[non_exhaustive]
pub enum InferenceError {
/// Expected and actual lengths of some array-shaped input differ.
MismatchedShape {
@@ -8,15 +9,35 @@ pub enum InferenceError {
expected: usize,
got: usize,
},
/// An `Outcome` of the wrong variant was supplied for the requested inference.
WrongOutcomeKind {
context: &'static str,
expected: &'static str,
got: &'static str,
},
/// A probability value is outside `[0, 1]`.
InvalidProbability { value: f64 },
/// A scalar parameter is outside its valid range.
InvalidParameter { name: &'static str, value: f64 },
/// An event contains tied teams, but the draw probability is zero.
///
/// A zero draw probability asserts that draws cannot occur, so a tied
/// result has no representable likelihood. Configure a positive `p_draw`
/// (via `HistoryBuilder::p_draw` or `GameOptions::p_draw`) to admit ties.
TieWithoutDrawProbability { teams: (usize, usize) },
/// Convergence exceeded `max_iter` without falling below `epsilon`.
ConvergenceFailed {
last_step: (f64, f64),
iterations: usize,
},
/// Inference produced a non-finite value (NaN or infinity).
///
/// Indicates numerical breakdown; the resulting skills are meaningless
/// and must not be treated as a converged estimate.
NonFiniteResult {
context: &'static str,
step: (f64, f64),
},
/// Negative precision: a Gaussian with `pi < 0` slipped into an API call.
NegativePrecision { pi: f64 },
}
@@ -31,9 +52,29 @@ impl fmt::Display for InferenceError {
} => {
write!(f, "{kind}: expected length {expected}, got {got}")
}
Self::WrongOutcomeKind {
context,
expected,
got,
} => {
write!(f, "{context}: expected {expected}, got {got}")
}
Self::InvalidProbability { value } => {
write!(f, "probability must be in [0, 1]; got {value}")
}
Self::TieWithoutDrawProbability { teams } => {
write!(
f,
"teams {} and {} are tied, but p_draw is 0.0; set a positive draw probability to admit ties",
teams.0, teams.1
)
}
Self::NonFiniteResult { context, step } => {
write!(
f,
"{context}: inference produced a non-finite result (step = {step:?})"
)
}
Self::InvalidParameter { name, value } => {
write!(f, "{name} is invalid: {value}")
}
+19 -9
View File
@@ -458,11 +458,18 @@ impl<T: Time, D: Drift<T>> Game<'_, T, D> {
let ranks = outcome
.as_ranks()
.ok_or(crate::InferenceError::MismatchedShape {
kind: "Game::ranked requires Outcome::Ranked",
expected: 0,
got: 0,
.ok_or(crate::InferenceError::WrongOutcomeKind {
context: "Game::ranked",
expected: "Outcome::Ranked",
got: "Outcome::Scored",
})?;
if options.p_draw == 0.0
&& let Some(tied) = crate::first_tied_pair(ranks)
{
return Err(crate::InferenceError::TieWithoutDrawProbability { teams: tied });
}
let max_rank = ranks.iter().copied().max().unwrap_or(0) as f64;
let result: Vec<f64> = ranks.iter().map(|&r| max_rank - r as f64).collect();
let teams_owned: Vec<Vec<Rating<T, D>>> = teams.iter().map(|t| t.to_vec()).collect();
@@ -497,10 +504,10 @@ impl<T: Time, D: Drift<T>> Game<'_, T, D> {
}
let scores = outcome
.as_scores()
.ok_or(crate::InferenceError::MismatchedShape {
kind: "Game::scored requires Outcome::Scored",
expected: 0,
got: 0,
.ok_or(crate::InferenceError::WrongOutcomeKind {
context: "Game::scored",
expected: "Outcome::Scored",
got: "Outcome::Ranked",
})?
.to_vec();
let teams_owned: Vec<Vec<Rating<T, D>>> = teams.iter().map(|t| t.to_vec()).collect();
@@ -1124,7 +1131,10 @@ mod tests {
&GameOptions::default(),
)
.unwrap_err();
assert!(matches!(err, crate::InferenceError::MismatchedShape { .. }));
assert!(matches!(
err,
crate::InferenceError::WrongOutcomeKind { .. }
));
}
#[test]
+53 -5
View File
@@ -220,6 +220,10 @@ impl<T: Time, D: Drift<T>, O: Observer<T>, K: Eq + Hash + Clone> History<T, D, O
fn iteration(&mut self) -> (f64, f64) {
let mut step = (0.0, 0.0);
if self.time_slices.is_empty() {
return step;
}
competitor::clean(self.agents.values_mut(), false);
for j in (0..self.time_slices.len() - 1).rev() {
@@ -435,6 +439,18 @@ impl<T: Time, D: Drift<T>, O: Observer<T>, K: Eq + Hash + Clone> History<T, D, O
use smallvec::SmallVec;
let opts = self.convergence;
if self.time_slices.is_empty() {
return Ok(ConvergenceReport {
iterations: 0,
final_step: (0.0, 0.0),
log_evidence: 0.0,
converged: true,
per_iteration_time: SmallVec::new(),
slices_skipped: 0,
});
}
let mut step = (f64::INFINITY, f64::INFINITY);
let mut i = 0;
let mut per_iter: SmallVec<[std::time::Duration; 32]> = SmallVec::new();
@@ -444,8 +460,24 @@ impl<T: Time, D: Drift<T>, O: Observer<T>, K: Eq + Hash + Clone> History<T, D, O
per_iter.push(t0.elapsed());
i += 1;
self.observer.on_iteration_end(i, step);
// A non-finite step means EP has broken down; further iterations
// cannot recover, and `tuple_gt` would read NaN as converged.
if !crate::step_is_finite(step) {
break;
}
}
let converged = !tuple_gt(step, opts.epsilon);
if !crate::step_is_finite(step) {
self.observer.on_converged(i, step, false);
return Err(InferenceError::NonFiniteResult {
context: "History::converge",
step,
});
}
let converged = crate::step_converged(step, opts.epsilon);
let log_evidence = self.log_evidence_internal(false, &[]);
self.observer.on_converged(i, step, converged);
Ok(ConvergenceReport {
@@ -498,6 +530,19 @@ impl<T: Time, D: Drift<T>, O: Observer<T>, K: Eq + Hash + Clone> History<T, D, O
});
}
// Chokepoint for tie validation: every ingestion route lands here,
// including `record_draw`, which builds its results directly rather
// than going through `Outcome`.
if self.p_draw == 0.0 {
for (event_results, kind) in results.iter().zip(kinds.iter()) {
if matches!(kind, EventKind::Ranked)
&& let Some(tied) = crate::first_tied_output(event_results)
{
return Err(InferenceError::TieWithoutDrawProbability { teams: tied });
}
}
}
competitor::clean(self.agents.values_mut(), true);
let mut this_agent = Vec::with_capacity(1024);
@@ -734,10 +779,13 @@ impl<T: Time, D: Drift<T>, O: Observer<T>, K: Eq + Hash + Clone> History<T, D, O
}
crate::Outcome::Scored { scores, sigma } => {
let resolved = sigma.unwrap_or(self.score_sigma);
debug_assert!(
resolved > 0.0,
"resolved score_sigma must be > 0.0 (got {resolved})"
);
if !(resolved > 0.0) {
return Err(InferenceError::InvalidParameter {
name: "score_sigma",
value: resolved,
});
}
kinds.push(EventKind::Scored {
score_sigma: resolved,
});
+50
View File
@@ -184,6 +184,56 @@ pub(crate) fn tuple_gt(t: (f64, f64), e: f64) -> bool {
t.0 > e || t.1 > e
}
/// Whether a convergence step is finite in both components.
///
/// A NaN step means EP broke down numerically. Because every comparison
/// against NaN is false, `tuple_gt` reads NaN as "below epsilon" — so
/// convergence checks must test finiteness explicitly rather than inferring
/// success from `!tuple_gt(..)`.
pub(crate) fn step_is_finite(t: (f64, f64)) -> bool {
t.0.is_finite() && t.1.is_finite()
}
/// Whether a step counts as converged: finite *and* within `epsilon`.
pub(crate) fn step_converged(t: (f64, f64), epsilon: f64) -> bool {
step_is_finite(t) && !tuple_gt(t, epsilon)
}
/// Indices of the first pair of teams sharing a rank, if any.
///
/// A tie is only representable when the draw probability is positive: with
/// `p_draw == 0.0` the truncation margin collapses to zero and the two-sided
/// tie update evaluates `0/0`. Callers use this to reject such events before
/// they reach inference.
pub(crate) fn first_tied_pair(ranks: &[u32]) -> Option<(usize, usize)> {
for (i, a) in ranks.iter().enumerate() {
for (j, b) in ranks.iter().enumerate().skip(i + 1) {
if a == b {
return Some((i, j));
}
}
}
None
}
/// As `first_tied_pair`, but over the engine's internal `f64` outputs.
///
/// Ranks reach the engine already converted to descending `f64` outputs, and
/// `Game` decides a tie by exact equality of those values — so this mirrors
/// the comparison inference itself performs.
pub(crate) fn first_tied_output(outputs: &[f64]) -> Option<(usize, usize)> {
for (i, a) in outputs.iter().enumerate() {
for (j, b) in outputs.iter().enumerate().skip(i + 1) {
if a == b {
return Some((i, j));
}
}
}
None
}
pub(crate) fn sort_time<T: Copy + Ord>(xs: &[T], reverse: bool) -> Vec<usize> {
let mut x: Vec<(usize, T)> = xs.iter().enumerate().map(|(i, &t)| (i, t)).collect();
+13 -5
View File
@@ -57,9 +57,11 @@ impl Outcome {
/// Explicit per-team continuous scores with a per-event noise override.
///
/// `sigma` must be `> 0.0`; debug-asserts otherwise.
/// `sigma` must be `> 0.0`. Constructing an `Outcome` with a non-positive
/// or NaN sigma is allowed; the value is rejected with
/// `InferenceError::InvalidParameter` when the event is ingested, so
/// callers get an error rather than a panic.
pub fn scores_with_sigma<I: IntoIterator<Item = f64>>(scores: I, sigma: f64) -> Self {
debug_assert!(sigma > 0.0, "score_sigma must be > 0.0 (got {sigma})");
Self::Scored {
scores: scores.into_iter().collect(),
sigma: Some(sigma),
@@ -169,9 +171,15 @@ mod tests {
}
}
/// Construction accepts any sigma; the value is validated at ingestion so
/// callers receive an `InferenceError` rather than a panic. See
/// `tests/degenerate_inputs.rs::scored_event_rejects_non_positive_sigma`.
#[test]
#[should_panic(expected = "score_sigma must be > 0.0")]
fn scores_with_sigma_rejects_zero() {
let _ = Outcome::scores_with_sigma([3.0, 1.0], 0.0);
fn scores_with_sigma_defers_validation_to_ingestion() {
let o = Outcome::scores_with_sigma([3.0, 1.0], 0.0);
match o {
Outcome::Scored { sigma, .. } => assert_eq!(sigma, Some(0.0)),
Outcome::Ranked(_) => panic!("expected Scored variant"),
}
}
}
+195
View File
@@ -0,0 +1,195 @@
//! Degenerate, boundary, and error-path coverage.
//!
//! These run in both debug and release: the defects they pin were all
//! guarded only by `debug_assert!`, so a debug-only suite never saw them.
use trueskill_tt::{
ConstantDrift, ConvergenceOptions, Game, GameOptions, Gaussian, History, InferenceError,
Outcome, Rating,
};
type R = Rating<i64, ConstantDrift>;
fn rating() -> R {
R::new(
Gaussian::from_ms(25.0, 25.0 / 3.0),
25.0 / 6.0,
ConstantDrift(25.0 / 300.0),
)
}
fn assert_finite(g: Gaussian, what: &str) {
assert!(
g.mu().is_finite() && g.sigma().is_finite(),
"{what} must be finite, got mu={} sigma={}",
g.mu(),
g.sigma()
);
}
#[test]
fn record_draw_without_draw_probability_is_rejected() {
let mut h = History::default();
let err = h.record_draw(&"a", &"b", 1).unwrap_err();
assert!(matches!(
err,
InferenceError::TieWithoutDrawProbability { .. }
));
}
#[test]
fn builder_draw_without_draw_probability_is_rejected() {
let mut h = History::default();
let err = h
.event(1)
.team(["a"])
.team(["b"])
.draw()
.commit()
.unwrap_err();
assert!(matches!(
err,
InferenceError::TieWithoutDrawProbability { .. }
));
}
#[test]
fn draw_with_positive_draw_probability_is_finite() {
let mut h = History::builder().p_draw(0.25).build();
h.record_draw(&"a", &"b", 1).unwrap();
let report = h.converge().unwrap();
assert_finite(h.current_skill("a").unwrap(), "drawn competitor skill");
assert_finite(h.current_skill("b").unwrap(), "drawn competitor skill");
assert!(report.log_evidence.is_finite());
assert!(report.converged);
}
#[test]
fn game_ranked_rejects_tie_without_draw_probability() {
let a = [rating()];
let b = [rating()];
let teams: Vec<&[R]> = vec![&a, &b];
let err = Game::ranked(&teams, Outcome::draw(2), &GameOptions::default()).unwrap_err();
assert!(matches!(
err,
InferenceError::TieWithoutDrawProbability { .. }
));
}
/// `Outcome::winner(w, n)` ties every loser, so any n >= 3 free-for-all hits
/// the tie path even though the caller never asked for a draw.
#[test]
fn winner_of_three_or_more_requires_draw_probability() {
let a = [rating()];
let b = [rating()];
let c = [rating()];
let teams: Vec<&[R]> = vec![&a, &b, &c];
let err = Game::ranked(&teams, Outcome::winner(0, 3), &GameOptions::default()).unwrap_err();
assert!(matches!(
err,
InferenceError::TieWithoutDrawProbability { .. }
));
let opts = GameOptions {
p_draw: 0.1,
..GameOptions::default()
};
let game = Game::ranked(&teams, Outcome::winner(0, 3), &opts).unwrap();
for team in game.posteriors() {
for skill in team {
assert_finite(skill, "3-team winner posterior");
}
}
}
#[test]
fn full_ranking_without_ties_needs_no_draw_probability() {
let a = [rating()];
let b = [rating()];
let c = [rating()];
let teams: Vec<&[R]> = vec![&a, &b, &c];
let game = Game::ranked(&teams, Outcome::ranking([0, 1, 2]), &GameOptions::default()).unwrap();
for team in game.posteriors() {
for skill in team {
assert_finite(skill, "strict ranking posterior");
}
}
}
#[test]
fn empty_history_converges_trivially() {
let mut h = History::default();
let report = h.converge().unwrap();
assert_eq!(report.iterations, 0);
assert!(report.converged);
}
#[test]
fn empty_event_stream_then_converge() {
let mut h = History::default();
h.add_events(std::iter::empty()).unwrap();
let report = h.converge().unwrap();
assert_eq!(report.iterations, 0);
}
#[test]
fn empty_history_queries_do_not_panic() {
let h = History::default();
assert!(h.learning_curves().is_empty());
assert!(h.learning_curve("nobody").is_empty());
assert!(h.current_skill("nobody").is_none());
}
#[test]
fn single_event_history_converges() {
let mut h = History::default();
h.record_winner(&"a", &"b", 1).unwrap();
let report = h.converge().unwrap();
assert!(report.converged);
assert_finite(h.current_skill("a").unwrap(), "single-event skill");
}
#[test]
fn scored_event_rejects_non_positive_sigma() {
let mut h = History::builder().score_sigma(2.0).build();
let err = h
.event(1)
.team(["a"])
.team(["b"])
.scores_with_sigma([3.0, 1.0], f64::NAN)
.commit()
.unwrap_err();
assert!(matches!(
err,
InferenceError::InvalidParameter {
name: "score_sigma",
..
}
));
}
#[test]
fn convergence_reports_are_finite_across_many_teams() {
let opts = GameOptions {
p_draw: 0.1,
convergence: ConvergenceOptions::default(),
..GameOptions::default()
};
let holders: Vec<[R; 1]> = (0..12).map(|_| [rating()]).collect();
let teams: Vec<&[R]> = holders.iter().map(|t| t.as_slice()).collect();
let game = Game::ranked(&teams, Outcome::ranking(0..12), &opts).unwrap();
assert!(
game.log_evidence().is_finite(),
"12-team log-evidence must be finite, got {}",
game.log_evidence()
);
for team in game.posteriors() {
for skill in team {
assert_finite(skill, "12-team posterior");
}
}
}